(This is a draft and truncated version - for final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals)
adjusted odds ratio (adjusted OR), see also odds ratio
As the name implies, the odds ratio is the ratio of the odds of presence of an antecedent in those with positive outcome to the odds in those with negative outcome. This ratio needs to be adjusted when the outcome is suspected to be affected by other factors. For example, presence of oral cancer in any case is affected by whether or not the person consumed smokeless tobacco for long duration. Let this be the antecedent of interest. But oral cancer can also occur or accelerate by insufficient intake of fruits and vegetables and presence of leukoplakia. When such concomitant variables, called covariates, are ignored altogether, what is obtained is called unadjusted OR. But when the analysis is geared to remove the influence of these covariates, adjusted OR is obtained. For example, you may have an unadjusted OR = 12.7 for smokeless tobacco in cases of oral cancer, and after adjustment of the covariates this may reduce to 8.9. The balance 3.8 can be attributed to the covariates in the model. Adjusted OR will decrease if the covariates tend to increase the incidence of disease and will increase if the covariates decrease the incidence.
The most common method for finding the adjusted OR is the logistic regression. In this method, the logistic coefficients are the logarithm of respective ORs. This regression can be run with several covariates in the logistic model. Each coefficient provides ln(OR) for that factor and this is automatically adjusted for the other covariates in the model. In our example, the coefficient of leukoplakia will provide OR adjusted for use of smokeless tobacco and insufficient intake of fruits and vegetables, and the coefficient of use of smokeless will provide OR adjusted for insufficient intake of fruits and vegetables and presence of leukoplakia. The same regression model can provide several adjusted ORs.
Adjusted OR is often misused. It is incorrect to interpret the adjusted OR as the net effect of the antecedent. Sometimes this is termed as the independent contribution of the antecedent. This term is also suspect. The adjustment is only for those covariates which are included in your model. There might be other factors that are not present in the model, including those that are unknown but may influence the outcome. Since no adjustment is done for these covariates, adjusted OR does not measure the net effect, nor even the independent effect. The best way to understand this is the effect adjusted for the other covariates in the model. In our example, 8.9 is the OR adjusted for intake of fruits and vegetables and presence of leukoplakia. This is not adjusted for other factors such as family history and exposure to sun, and this is not the net effect either. ... ...
For final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals
Concise Encyclopedia of Biostatistics for Medical Professionals)
adjusted odds ratio (adjusted OR), see also odds ratio
As the name implies, the odds ratio is the ratio of the odds of presence of an antecedent in those with positive outcome to the odds in those with negative outcome. This ratio needs to be adjusted when the outcome is suspected to be affected by other factors. For example, presence of oral cancer in any case is affected by whether or not the person consumed smokeless tobacco for long duration. Let this be the antecedent of interest. But oral cancer can also occur or accelerate by insufficient intake of fruits and vegetables and presence of leukoplakia. When such concomitant variables, called covariates, are ignored altogether, what is obtained is called unadjusted OR. But when the analysis is geared to remove the influence of these covariates, adjusted OR is obtained. For example, you may have an unadjusted OR = 12.7 for smokeless tobacco in cases of oral cancer, and after adjustment of the covariates this may reduce to 8.9. The balance 3.8 can be attributed to the covariates in the model. Adjusted OR will decrease if the covariates tend to increase the incidence of disease and will increase if the covariates decrease the incidence.
The most common method for finding the adjusted OR is the logistic regression. In this method, the logistic coefficients are the logarithm of respective ORs. This regression can be run with several covariates in the logistic model. Each coefficient provides ln(OR) for that factor and this is automatically adjusted for the other covariates in the model. In our example, the coefficient of leukoplakia will provide OR adjusted for use of smokeless tobacco and insufficient intake of fruits and vegetables, and the coefficient of use of smokeless will provide OR adjusted for insufficient intake of fruits and vegetables and presence of leukoplakia. The same regression model can provide several adjusted ORs.
Adjusted OR is often misused. It is incorrect to interpret the adjusted OR as the net effect of the antecedent. Sometimes this is termed as the independent contribution of the antecedent. This term is also suspect. The adjustment is only for those covariates which are included in your model. There might be other factors that are not present in the model, including those that are unknown but may influence the outcome. Since no adjustment is done for these covariates, adjusted OR does not measure the net effect, nor even the independent effect. The best way to understand this is the effect adjusted for the other covariates in the model. In our example, 8.9 is the OR adjusted for intake of fruits and vegetables and presence of leukoplakia. This is not adjusted for other factors such as family history and exposure to sun, and this is not the net effect either. ... ...
For final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals